Abstract: Non-point source pollution modeling is the most direct and effective way to quantify the pollution of this type. Certain errors generated in the simulation procedure, such as the one in nonpoint source, model structure, model parameters, or data sources, usually cause a large uncertainty in model outputs. This paper summarizes the structure, advantages, and limitations of the representative models of nonpoint source pollution over a watershed, and examines the uncertainties in their calculations, together with an overview on the analysis methods of the uncertainties. From these models’ previous application and existing problems, we suggest that in future modeling accuracy should be improved via adopting a correct procedure: to explore further the non-point source pollution mechanism, and then develop effective reliability models applicable to different data conditions, and combine with soil and water conservation, among other engineering measures. Most previous studies on such uncertainties focused on the uncertainty analysis of single factors, but other analysis methods should also be adopted in future to optimize model structure and find out the best structure for specific problems. And uncertainty methods for multiple systems could be systematically studied to achieve a comprehensive assessment of simulation results.